In most East Asian languages, language-specific characters in code page strings (that is, Natural format A) are represented by 2 bytes (the so-called double-byte characters) and ASCII characters (EBCDIC on mainframes) are represented by 1 byte. Thus, a code pages string consists of characters with different lengths: some have 1 byte and others have 2 bytes.
Natural provides a basic support for double-byte characters. On Windows,
this support is activated when both the Natural default code page and the
Windows system code page are defined as double-byte code pages. If Natural does
not define a specific code page, it is sufficient when a double-byte Windows
system code page has been defined. On UNIX and OpenVMS, the support for
double-byte characters is activated when the Natural default code page is a
double-byte code page. On mainframes, the profile parameter
CP
must be set to an EBCDIC MBCS code page, for example IBM-942.
When double-byte character support is enabled, Natural assures for all string manipulations that a double-byte character is treated as a unit. This is essential for keeping the meaning of a string.
If a single leading or trailing byte of a double-byte character is left
over after the manipulation of a variable of format A (for example, after
extracting a substring with the SUBSTRING option), this byte is
replaced with a blank character.
For the example below, the code page Shift_JIS is selected. Variable
#A contains a string which consists of four characters. The first
and last character is the double-byte character "FULL WIDTH LATIN
SMALL LETTER B" which is represented in code page Shift_JIS by
the byte sequence H'8282'. The second and third character is the
single byte character "LATIN SMALL LETTER A" which
is represented by one byte H'61'. Thus, the hexadecimal
representation of the full string is H'828261618282'.
DEFINE DATA LOCAL 1 #A (A10) END-DEFINE #A := 'aa
' REPLACE 'a' WRITE #A #A (EM=H(6)) END
Without double-byte character support the output of the above program is as follows:
Page 1 07-02-07 17:22:09
This is the result of not having treated the character "" (H'8282' in code page Shift_JIS) as one unit. The
trailing byte of this character and the following character
"a" (H'61') are falsely interpreted as
the double-byte character "" (H'8261' in code page Shift_JIS).
With double-byte character support, the output of the program is as expected:
Page 1 07-02-07 17:22:09
